summarize_conversation

Generate concise summaries of Carbon Voice conversations to extract key information and insights from message exchanges.

Instructions

Summarize a conversation.

Input Schema

TableJSON Schema

Name	Required	Description	Default
`conversation_id`	Yes
`prompt_id`	Yes
`message_ids`	No
`language`	No
`start_date`	No
`end_date`	No
`limit`	No

Implementation Reference

src/server.ts:487-529 (registration)

Full registration of the 'summarize_conversation' tool, including the inline anonymous handler function that implements the tool's logic by fetching messages if necessary and calling the AI response API to generate a summary.

server.registerTool(
  'summarize_conversation',
  {
    description: 'Summarize a conversation.',
    inputSchema: summarizeConversationParams.shape,
    annotations: {
      readOnlyHint: false,
      destructiveHint: false,
    },
  },
  async (
    args: SummarizeConversationParams,
    { authInfo },
  ): Promise<McpToolResponse> => {
    try {
      let message_ids: string[] = args.message_ids || [];

      // If no message ids are provided, get couple of messages from the conversation
      if (!args.message_ids) {
        const messages = await simplifiedApi.listMessages(
          args,
          setCarbonVoiceAuthHeader(authInfo?.token),
        );
        message_ids = messages.results?.map((message) => message.id) || [];
      }

      const aiResponse = await simplifiedApi.aIResponseControllerCreateResponse(
        {
          prompt_id: args.prompt_id,
          message_ids: message_ids,
          channel_id: args.conversation_id,
          language: args.language,
        },
        setCarbonVoiceAuthHeader(authInfo?.token),
      );

      return formatToMCPToolResponse(aiResponse);
    } catch (error) {
      logger.error('Error summarizing conversation:', { error });
      return formatToMCPToolResponse(error);
    }
  },
);

src/server.ts:497-528 (handler)

The anonymous async handler function that executes the core logic of the summarize_conversation tool: conditionally lists recent messages from the conversation, then invokes the AI response controller with a specific prompt to generate the summary.

async (
  args: SummarizeConversationParams,
  { authInfo },
): Promise<McpToolResponse> => {
  try {
    let message_ids: string[] = args.message_ids || [];

    // If no message ids are provided, get couple of messages from the conversation
    if (!args.message_ids) {
      const messages = await simplifiedApi.listMessages(
        args,
        setCarbonVoiceAuthHeader(authInfo?.token),
      );
      message_ids = messages.results?.map((message) => message.id) || [];
    }

    const aiResponse = await simplifiedApi.aIResponseControllerCreateResponse(
      {
        prompt_id: args.prompt_id,
        message_ids: message_ids,
        channel_id: args.conversation_id,
        language: args.language,
      },
      setCarbonVoiceAuthHeader(authInfo?.token),
    );

    return formatToMCPToolResponse(aiResponse);
  } catch (error) {
    logger.error('Error summarizing conversation:', { error });
    return formatToMCPToolResponse(error);
  }
},

src/schemas/conversation.ts:3-11 (schema)

Zod schema defining the input parameters for the summarize_conversation tool, including conversation_id, prompt_id, optional message_ids, language, date ranges, and limit.

export const summarizeConversationParams = z.object({
  conversation_id: z.string().nonempty(),
  prompt_id: z.string().nonempty(),
  message_ids: z.array(z.string()).optional(),
  language: z.string().optional(),
  start_date: z.string().datetime().optional(),
  end_date: z.string().datetime().optional(),
  limit: z.number().optional().default(50),
});

src/interfaces/conversation.interface.ts:8-10 (schema)
TypeScript type definition for SummarizeConversationParams, inferred from the Zod schema for type safety in the handler.
```
export type SummarizeConversationParams = z.infer<
  typeof summarizeConversationParams
>;
```

Carbon Voice